1#ifndef _FS_CEPH_OSDMAP_H
2#define _FS_CEPH_OSDMAP_H
3
4#include <linux/rbtree.h>
5#include <linux/ceph/types.h>
6#include <linux/ceph/decode.h>
7#include <linux/ceph/ceph_fs.h>
8#include <linux/crush/crush.h>
9
10/*
11 * The osd map describes the current membership of the osd cluster and
12 * specifies the mapping of objects to placement groups and placement
13 * groups to (sets of) osds.  That is, it completely specifies the
14 * (desired) distribution of all data objects in the system at some
15 * point in time.
16 *
17 * Each map version is identified by an epoch, which increases monotonically.
18 *
19 * The map can be updated either via an incremental map (diff) describing
20 * the change between two successive epochs, or as a fully encoded map.
21 */
22struct ceph_pg {
23	uint64_t pool;
24	uint32_t seed;
25};
26
27#define CEPH_POOL_FLAG_HASHPSPOOL  1
28
29struct ceph_pg_pool_info {
30	struct rb_node node;
31	s64 id;
32	u8 type;
33	u8 size;
34	u8 crush_ruleset;
35	u8 object_hash;
36	u32 pg_num, pgp_num;
37	int pg_num_mask, pgp_num_mask;
38	s64 read_tier;
39	s64 write_tier; /* wins for read+write ops */
40	u64 flags;
41	char *name;
42};
43
44static inline bool ceph_can_shift_osds(struct ceph_pg_pool_info *pool)
45{
46	switch (pool->type) {
47	case CEPH_POOL_TYPE_REP:
48		return true;
49	case CEPH_POOL_TYPE_EC:
50		return false;
51	default:
52		BUG_ON(1);
53	}
54}
55
56struct ceph_object_locator {
57	s64 pool;
58};
59
60/*
61 * Maximum supported by kernel client object name length
62 *
63 * (probably outdated: must be >= RBD_MAX_MD_NAME_LEN -- currently 100)
64 */
65#define CEPH_MAX_OID_NAME_LEN 100
66
67struct ceph_object_id {
68	char name[CEPH_MAX_OID_NAME_LEN];
69	int name_len;
70};
71
72struct ceph_pg_mapping {
73	struct rb_node node;
74	struct ceph_pg pgid;
75
76	union {
77		struct {
78			int len;
79			int osds[];
80		} pg_temp;
81		struct {
82			int osd;
83		} primary_temp;
84	};
85};
86
87struct ceph_osdmap {
88	struct ceph_fsid fsid;
89	u32 epoch;
90	u32 mkfs_epoch;
91	struct ceph_timespec created, modified;
92
93	u32 flags;         /* CEPH_OSDMAP_* */
94
95	u32 max_osd;       /* size of osd_state, _offload, _addr arrays */
96	u8 *osd_state;     /* CEPH_OSD_* */
97	u32 *osd_weight;   /* 0 = failed, 0x10000 = 100% normal */
98	struct ceph_entity_addr *osd_addr;
99
100	struct rb_root pg_temp;
101	struct rb_root primary_temp;
102
103	u32 *osd_primary_affinity;
104
105	struct rb_root pg_pools;
106	u32 pool_max;
107
108	/* the CRUSH map specifies the mapping of placement groups to
109	 * the list of osds that store+replicate them. */
110	struct crush_map *crush;
111
112	struct mutex crush_scratch_mutex;
113	int crush_scratch_ary[CEPH_PG_MAX_SIZE * 3];
114};
115
116static inline void ceph_oid_set_name(struct ceph_object_id *oid,
117				     const char *name)
118{
119	int len;
120
121	len = strlen(name);
122	if (len > sizeof(oid->name)) {
123		WARN(1, "ceph_oid_set_name '%s' len %d vs %zu, truncating\n",
124		     name, len, sizeof(oid->name));
125		len = sizeof(oid->name);
126	}
127
128	memcpy(oid->name, name, len);
129	oid->name_len = len;
130}
131
132static inline void ceph_oid_copy(struct ceph_object_id *dest,
133				 struct ceph_object_id *src)
134{
135	BUG_ON(src->name_len > sizeof(dest->name));
136	memcpy(dest->name, src->name, src->name_len);
137	dest->name_len = src->name_len;
138}
139
140static inline int ceph_osd_exists(struct ceph_osdmap *map, int osd)
141{
142	return osd >= 0 && osd < map->max_osd &&
143	       (map->osd_state[osd] & CEPH_OSD_EXISTS);
144}
145
146static inline int ceph_osd_is_up(struct ceph_osdmap *map, int osd)
147{
148	return ceph_osd_exists(map, osd) &&
149	       (map->osd_state[osd] & CEPH_OSD_UP);
150}
151
152static inline int ceph_osd_is_down(struct ceph_osdmap *map, int osd)
153{
154	return !ceph_osd_is_up(map, osd);
155}
156
157static inline bool ceph_osdmap_flag(struct ceph_osdmap *map, int flag)
158{
159	return map && (map->flags & flag);
160}
161
162extern char *ceph_osdmap_state_str(char *str, int len, int state);
163extern u32 ceph_get_primary_affinity(struct ceph_osdmap *map, int osd);
164
165static inline struct ceph_entity_addr *ceph_osd_addr(struct ceph_osdmap *map,
166						     int osd)
167{
168	if (osd >= map->max_osd)
169		return NULL;
170	return &map->osd_addr[osd];
171}
172
173static inline int ceph_decode_pgid(void **p, void *end, struct ceph_pg *pgid)
174{
175	__u8 version;
176
177	if (!ceph_has_room(p, end, 1 + 8 + 4 + 4)) {
178		pr_warn("incomplete pg encoding\n");
179		return -EINVAL;
180	}
181	version = ceph_decode_8(p);
182	if (version > 1) {
183		pr_warn("do not understand pg encoding %d > 1\n",
184			(int)version);
185		return -EINVAL;
186	}
187
188	pgid->pool = ceph_decode_64(p);
189	pgid->seed = ceph_decode_32(p);
190	*p += 4;	/* skip deprecated preferred value */
191
192	return 0;
193}
194
195extern struct ceph_osdmap *ceph_osdmap_decode(void **p, void *end);
196extern struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
197					    struct ceph_osdmap *map,
198					    struct ceph_messenger *msgr);
199extern void ceph_osdmap_destroy(struct ceph_osdmap *map);
200
201/* calculate mapping of a file extent to an object */
202extern int ceph_calc_file_object_mapping(struct ceph_file_layout *layout,
203					 u64 off, u64 len,
204					 u64 *bno, u64 *oxoff, u64 *oxlen);
205
206/* calculate mapping of object to a placement group */
207extern int ceph_oloc_oid_to_pg(struct ceph_osdmap *osdmap,
208			       struct ceph_object_locator *oloc,
209			       struct ceph_object_id *oid,
210			       struct ceph_pg *pg_out);
211
212extern int ceph_calc_pg_acting(struct ceph_osdmap *osdmap,
213			       struct ceph_pg pgid,
214			       int *osds, int *primary);
215extern int ceph_calc_pg_primary(struct ceph_osdmap *osdmap,
216				struct ceph_pg pgid);
217
218extern struct ceph_pg_pool_info *ceph_pg_pool_by_id(struct ceph_osdmap *map,
219						    u64 id);
220
221extern const char *ceph_pg_pool_name_by_id(struct ceph_osdmap *map, u64 id);
222extern int ceph_pg_poolid_by_name(struct ceph_osdmap *map, const char *name);
223
224#endif
225